MLCommons®, the open engineering consortium behind the industry-standard MLPerf® benchmarks, today announced the release of MLPerf Client v1.6, the latest version of its benchmark suite for evaluating AI performance on personal computers.
MLPerf Client measures how well PCs—from laptops and desktops to workstations—run local AI workloads such as large language models (LLMs). By simulating real generative AI tasks, including text summarization, content creation, and code analysis, the benchmark provides clear, standardized metrics like responsiveness and throughput.
Version 1.6 continues to optimize the benchmark experience, updating key software components and improving usability and performance.
What's New in MLPerf Client v1.6
MLPerf Client v1.6 enhances acceleration support by updating core AI runtimes and frameworks. This includes new versions of Windows ML and llama.cpp, as well as the latest runtime optimizations from independent hardware vendors on the Windows platform. On Apple platforms, updates to MLX with Metal and llama.cpp with Metal further improve performance and compatibility on macOS and iPad systems.
The release also introduces several improvements to the graphical user interface: optimized startup performance to reduce application loading time, a new progress bar providing clear feedback during initialization, and internal architectural refactoring to enhance overall stability during benchmark runs.
To simplify repetitive testing workflows, MLPerf Client v1.6 adds an option to disable the download confirmation prompt when starting a benchmark. This allows users to efficiently run multiple test batches with a single click, boosting productivity for reviewers and developers.
These updates further solidify MLPerf Client as a reliable, easy-to-use benchmark for evaluating AI performance across client systems.
MLPerf Client is developed collaboratively by leading technology companies, including AMD, Intel, Microsoft, NVIDIA, Qualcomm Technologies, Inc., and major PC OEMs. The benchmark is fully open source and free; the source code is available for review and contribution on the MLCommons GitHub repository.
For downloads, documentation, and more information, visit mlcommons.org/benchmarks/client. The GUI version is also available on the iOS and Mac App Store, and will be released on Steam in the coming days.
About MLCommons
MLCommons is an open engineering consortium dedicated to making machine learning accessible to everyone. The organization develops leading benchmarks, datasets, and best practices covering cloud, data center, edge, and client AI systems. Its MLPerf benchmark suite is widely recognized as the industry standard for measuring machine learning performance.
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接